Topological and Spatial Heterogeneity Measures in Medical Imaging: Methods, Validation and Clinical Value

Quantitative medical imaging has traditionally focused on intensity, shape and texture, while the spatial organization of tissues and lesions has received comparatively less attention. Topological and spatial descriptors - including persistent homology, Betti numbers, Minkowski functionals, habitat analysis and graph-based representations- can characterize how regions connect, branch, enclose and fragment across imaging data. These measures are increasingly incorporated into machine learning pipelines for diagnosis, prognosis, treatment-response assessment and longitudinal monitoring. However, their clinical value remains uncertain because they are calculated using diverse conventions, dimensional representations, segmentation strategies and imaging resolutions. Important questions concern their repeatability across scanners and protocols, sensitivity to segmentation errors, incremental value over established radiomic features and biological interpretability. Recent studies suggest that topology-aware and spatially resolved features may capture complementary information about tumor architecture, vascular organization and tissue heterogeneity, but findings are difficult to compare because of inconsistent preprocessing and limited external validation. Standardized methodologies, transparent benchmarking and clinically grounded validation are therefore needed to determine whether these measures can support reliable radiological decision-making.This Research Topic aims to advance the development, validation and clinical translation of topological and spatial heterogeneity measures in medical imaging. It will address how these descriptors should be defined and implemented, which imaging and segmentation conditions influence their reliability, and whether they provide independent and clinically meaningful information beyond conventional intensity, shape and texture features. Contributions may also examine how topology-aware representations can be integrated into artificial intelligence models while retaining interpretability and supporting reproducible research. Studies should clarify the relationship between imaging-derived measures, underlying biology, pathology and patient outcomes, and should report validation strategies appropriate to their design, with multi-center or external validation strongly encouraged.To gather further insights into topological and spatial descriptors derived from clinical or pre-clinical radiological imaging, this Research Topic will focus on methods with demonstrated relevance to imaging applications and will not include primarily mathematical studies lacking such an application. We welcome articles addressing, but not limited to, the following themes:• Topological descriptors of tumors and tissues from CT, MRI, PET and multimodal imaging, including persistent homology, Betti numbers and Minkowski functionals• Graph-, network- and connectivity-based representations of anatomical and pathological structures• Two-dimensional and three-dimensional formulations, including their respective advantages, limitations and clinical implications• Habitat imaging, subregion analysis, clustering-based partitioning and other measures of intratumoural or intra-organ heterogeneity• Topology-preserving image registration and segmentation, including topological constraints and the effects of segmentation quality on downstream measurements• Benchmarking against conventional intensity, shape and texture radiomics, including assessments of independent predictive value• Integration of topological and spatial descriptors into machine learning and deep learning models for diagnosis, prognosis, treatment response and follow-up• Repeatability, reproducibility and robustness across scanners, acquisition protocols, resolutions and observers• Harmonization, domain adaptation and standardization strategies for multi-center imaging studies• Explainability and biological interpretation of topological measures in relation to pathology, disease mechanisms and clinical outcomes• Open-source software, reproducible workflows, benchmark studies and reference datasets• External, prospective or multi-center validation of imaging biomarkers and artificial intelligence modelsWe welcome Original Research, Methods, Brief Research Reports, Reviews, Mini Reviews, and Hypothesis and Theory articles. Purely computational analyses of public datasets should include validation appropriate to the study design and journal policy. Topic editors: Zhichao Zuo, Li Li, Yang Liu, Jinhong Zhao Source: https://www.frontiersin.org/research-topics/85345/topological-and-spatial-heterogeneity-measures-in-medical-imaging-methods-validation-and-clinical-value